Metacommunity analyses show an increase in ecological specialisation throughout the Ediacaran period.

The first animals appear during the late Ediacaran (572 to 541 Ma); an initial diversity increase was followed reduction in diversity, often interpreted as catastrophic mass extinction. We investigate Ediacaran ecosystem structure changes over this time period using the "Elements of Metacommunity Structure" framework to assess whether this diversity reduction in the Nama was likely caused by an external mass extinction, or internal metacommunity restructuring. The oldest metacommunity was characterised by taxa with wide environmental tolerances, and limited specialisation or intertaxa associations. Structuring increased in the second oldest metacommunity, with groups of taxa sharing synchronous responses to environmental gradients, aggregating into distinct communities. This pattern strengthened in the youngest metacommunity, with communities showing strong environmental segregation and depth structure. Thus, metacommunity structure increased in complexity, with increased specialisation and resulting in competitive exclusion, not a catastrophic environmental disaster, leading to diversity loss in the terminal Ediacaran. These results reveal that the complex eco-evolutionary dynamics associated with Cambrian diversification were established in the Ediacaran.

Introduction

One of the most dramatic events in the history of Earth is the sudden appearance of animals in the fossil record during the Ediacaran period (635 to 541 Ma), after billions of years of microbial life [1-3]. Ediacaran anatomies are particularly difficult to compare to modern phyla, which has hampered our understanding of Ediacaran evolution and how Ediacaran organisms relate to the Cambrian Explosion and extant animal phyla [4]. Patterns of taxonomic, morphological, and ecospace diversity change dramatically during the Ediacaran [5,6], which has led to the suggestion of several evolutionary radiations, corresponding to the Avalon, White Sea, and Nama assemblages [1,7-9]. These 3 assemblages consist of groupings of communities that occupy partially overlapping temporal intervals and water depths, with no significant litho-taphonomic or biogeographic influence [7,8,10]. The oldest assemblage, the Avalon (575 to 565 Ma), exhibits relatively limited ecological and morphological diversity [5,6], with only limited palaeoenvironmental influence on its composition and taxa interactions [11-14]. The White Sea assemblage (558 to 550 Ma) shows a large increase in morphological diversity, including putative bilaterians [5], in tandem with a greater ecological diversity that includes the appearance of grazing, herbivory, and widespread motility [15,16]. These innovations are coupled to the development of dense communities with high community heterogeneity between environments [16,17] and increased taxa sensitivity to fine-scale environment [12,18]. The Nama assemblage (549 to 543 Ma) includes the oldest biomineralising taxa and records a decrease in taxonomic diversity [5,19-21]. This reduction in taxonomic diversity, sometimes referred to as the “diversity drop,” has been suggested to correspond to a post-White Sea extinction around 550 Ma, which eliminated the majority of Ediacaran soft-bodied organisms [9,22-24]. This diversity drop has been suggested to be caused by either an environmental driven catastrophic environmental extinction or biotic replacement driven extinction [9,21,23,24]. Recent work has shown that a biotic replacement driven extinction, whereby mobile metazoans outcompeted soft-bodied Ediacaran organisms through bioturbation and ecosystem engineering, is unlikely, due in part to prolonged co-occurrence of trace fossils with soft-bodied biota [1,25]. Other currently unknown and/or unpreserved intrinsic causes behind a biotic replacement model cannot be excluded at the moment.

Previous studies have focused primarily on defining the different assemblages and what the underlying factors behind the different assemblages [7,8,10], looking at taxonomic and morphological diversity between assemblages [5,26] with little investigation of how the ecological structure within the assemblages differs. The network structures of the co-occurrence of Ediacaran body fossil and trace fossil taxa were compared by Muscente and colleagues [9] who found compartmentalisation of the assemblages within the total Ediacaran network. However, the prior cluster analyses of [7,8,10] and network analyses of [9] have not assessed the relative frequency of taxa co-occurrences within assemblages; i.e., whether they were statistically different to what would be expected by random chance, nor compared the ecological structure within each assemblage to known ecological models.

In this study, we will investigate the structural attributes within these assemblages using 3 analyses that have not previous been used to investigate Ediacaran macroecology. We used presence–absence data encompassing 86 Ediacaran localities and 124 taxa, with paleoenvironment, depth, lithology, time, and assemblage data from [8,24] (S1 Fig). Ediacaran fossils are commonly found preserved in situ, so their bedding planes (the rock surfaces that preserve the fossils) preserve near-compete censuses of the communities [15,27]. This exceptional preservation means that ecological analyses normally reserved for modern communities can be applied (e.g., [12,28]).

First, we will use the “Elements of Metacommunity Structure” (EMS) framework to investigate emergent properties of groups of connected communities that may arise from taxa interactions, dispersal, environmental filtering, and the interaction of these factors [29-31] (Fig 1). Most fossil metacommunities do not fulfil the requirements of random sampling that would be needed to analyse them with such an ecological framework. However due to their exceptional preservation, the Ediacaran metacommunities are an exceptional census of the benthic assemblages present at the time, making them amenable to be analysed within the EMS framework. EMS does not assume even dispersal across all sites, with intermediate levels of disturbance associated with the highest levels of filtering of community by biotic and abiotic factors [29], and dispersal limitation associated with negative turnover [32]. Ediacaran communities vary in how much they are separated in time and space, from ecological to geological time scales [8,9,13], and their organisms have been shown to have large dispersal ranges based on reproductive mode [28,33,34] and species occurrence over large space and time scales [35]. Because the connectivity of these Ediacaran communities via dispersal has been established, here we define metacommunities as sets of fossil localities (communities), which are connected by the dispersal of many species [29]. The EMS framework is a hierarchical analysis that identifies properties in site-by-taxa presence/absence matrices, which are related to the underlying processes shaping taxa distributions [31], but to date has limited application to the fossil record [36]. Three metacommunity metrics are calculated to determine the structure: coherence, turnover, and boundary clumping [29-31], which are hierarchical rather than independent of each other. The values and statistical significance of these metrics determine where the metacommunity fits within the 14 different metacommunity types within the EMS framework (Fig 1), with different metric combinations indicating different underlying processes behind the metacommunity structure. To determine whether an observed metric score differs significantly from random, we computed the z-score, which measures its distance from the mean of the randomisations (simulation mean) as the number of standard deviations (thus making it comparable across metrics with difference units). If the z-score is negative, the observed value is smaller than the simulated mean; if it is positive, then it is greater than the simulated mean; z ≥ │3│ indicates a significant deviation.

Fig 1

Idealised metacommunity structures adapted from [30].

These graphs show taxa abundance patterns in idealised metacommunities of several taxa (represented by different colours), which respond to a latent environmental gradient (they exhibit significant positive coherence). The first step of the analyses (START) is to determine whether the metacommunity exhibits positive, negative, or random coherence. Random coherence corresponds to NS metacommunity structure; negative coherence is a checkerboard pattern [45], so significant mutual exclusivity between species and sites. Positive coherence indicates mutual co-occurring taxa associations, and there are several different possible models. For positive coherence, turnover and boundary clumping are calculated to determine the type of metacommunity structure. Nonsignificant turnover corresponds to quasi structures. These EMF analyses enable the structure of metacommunities to be grouped into one of 14 models: (1) random; (2) checkerboard; (3–5) nested clumped, random, and hyperdispersed; (6–8) Clementsian, Gleasonian, and evenly spaced; (9–10) Quasi nester clumped and hyperdispersed; and (11–12) quasi Clementsian and evenly spaced. See S1 Table. EMF, Elements of Metacommunity Framework; NS, no significant.

Coherence is a measure of the extent to which all the taxa respond to the same environmental gradient, where this gradient may result from the interplay of several biotic and abiotic factors that differ between sites [37]. Coherence is positive when the taxa in the site-by-taxa matrix all respond to the same environmental gradient. Most extant well-sampled metacommunities display significant positive coherence due to similarities in evolutionary history, ecological preferences, or life history trade-offs within communities [37]. A significant negative coherent site-by-taxa matrix reflects a high number of mutually exclusive taxa pairs creating checkerboard patterns [7,9,10]. These checkerboard patterns do not have further underlying structure (in contrast to positive coherence patterns; Fig 1), as there is no discernible gradient to which all the taxa respond. Negative co-occurrences and significant segregation/checkerboard patterns can be formed from strong competition, grazing/herbivory, or strongly nonoverlapping niches, all of which form similar metacommunity patterns due the presence of mutually exclusive pairs of taxa [29,31,37]. A nonsignificant coherence reflects no significant metacommunity structuring (Fig 1). For metacommunities that have positive coherence, the turnover metric tests the amount of taxa replacement between sites [38]. If taxa ranges are nested within each other, there is less turnover than expected by chance along the gradient (significantly negative). If there are more differences in site taxa composition along the gradient than expected by chance, turnover is significantly positive and the structure is nonnested. Nonsignificant turnover indicates a weaker structuring mechanism, termed a quasi structure (Fig 1) [30]. Quasi structures have the same fundamental characteristics as the idealised structures, but because range turnover is not significantly different from random, it is likely that the underlying structuring mechanisms are weaker than those for which turnover is significant. The final metric, boundary clumping measures the extent to which taxa range limits cluster at the same sites across the environmental gradient [37]. The range limits can be clumped (significant positive), hyperdispersed (significant negative), or random (nonsignificant).

Positive coherence and negative turnover result in nested metacommunities with taxa-poor sites being predictable subsets of taxa-rich sites, implying that species are dispersal limited [39]. Nested metacommunities have been shown to be associated with a low degree of spatial connectivity and environmental variation [39] and have been shown to govern postextinction dynamics [40].

Clumped species boundaries tend to be associated with the transitions between different biomes, where 2 biological communities mix, in contrast to hyperdispersed species loss where species loss is evenly distributed across the range [30]. Positive coherence and turnover with hyperdispersed (negative) boundary clumping describes an evenly spaced metacommunity (Fig 1). Where coherence, turnover and boundary clumping are all positive, the metacommunity is classed as Clementsian (Fig 1), where groups of taxa with similar range boundaries co-occur and respond in a similar way to environment gradients [37,41]. Taxa within Clementsian metacommunities respond synchronously to environmental gradients, suggesting physiological or evolutionary trade-offs associated with environmental thresholds [42], tend to result from high levels of environmental variability and spatial connectivity [39], and are found to be the most common (e.g., [32]). When coherence and turnover are positive but there is no significant boundary clumping, the metacommunity is described as Gleasonian (Fig 1) where each taxon reacts individualistically to environmental gradients [30].

Secondly, we used Spearman rank correlations to test whether within-assemblage community composition is correlated with depth. The ordering of the sites was given by the ordination output from the EMS analyses (Fig 2), which is produced by reciprocal averaging, a type of correspondence analysis that ordinates the sites (y-axis of Fig 2) based on their species composition (x-axis of Fig 2) [31]. This ordering groups the sites together with similar community composition, and we can see from Fig 2 that the assemblages (indicated by different colours) are grouped together and that the depths (shown alongside the y-axis) show a correspondence with these assemblages, with the Avalon sites deeper, then increased shallowing up the y-axis and ordination with the Nama sites being the most shallow. This first-axis ranking of the sites was used to test whether there was a significant association with depth.

Fig 2

Ordinated data table.

The assemblage and palaeoenvironment for each locality are given on the left. Sites are ranked based on reciprocal averaging ordination. Right plot shows the incidence matrix for taxa (columns) for all sites (rows) along the inferred environmental gradient after ordination. Ordination was calculated according to occurrence resulting from the overall metacommunity analysis. The presence of a taxon is given by a coloured square, absence by white. Doushantuo and indeterminate sites were excluded from the assemblage-level analyses. Depth index indicates the relative depth of the locality (cf., [8]), as determined by palaeoenvironment. The data underlying this Figure can be found in 10.6084/m9.figshare.13664105"https://doi.org/10.5061/dryad.1mh30 and in S1 Data.

Thirdly, we will test to determine which pairwise taxa co-occurrences are significantly nonrandom, and whether any nonrandom co-occurrences are positive or negative. We use a combinatorics approach to test whether species pairs are randomly distributed among sites [43]. If co-occurrences are significantly nonrandom, this suggests a shared underlying ecological or evolutionary process. While the interpretation of co-occurrence data is complicated because co-occurrence does not necessarily correspond to interaction [44], here we interpret pairwise correlations (or associations) to within the wider EMS framework where co-occurrences are not taken necessarily as direct taxa interactions but could also indicate taxa environmental associations and/or disassociations.

Based on the literature, we can make predictions about how we may expect metacommunity structure to change throughout these Ediacaran assemblages. We predict that the increase in taxonomic and morphological diversity between the Avalon and White Sea assemblages [5] is reflected in more ecological complexity in terms of increased taxa co-occurrences. We predict that the total set of Ediacaran data exhibits strong metacommunity structure that reflects the previously recovered assemblages [7-9] and that the influence of environmental gradients increases between the Avalon and White Sea assemblages [18]. Finally, we will use these analyses to test between 3 hypotheses relating to the underlying causes behind the White Sea–Nama drop in taxonomic diversity: (1) Null hypothesis: Changes in taxonomic diversity are not present or are not detectable; (2) External mass extinction: test whether there is evidence of a catastrophic extinction event between the White Sea and Nama [9,22,46]. Such an event would lead to negative turnover, so a nested metacommunity structure ([40]; or 3) Internal restructuring: increased ecological complexity via co-occurrences and strong metacommunity leading to stronger niche partitioning.

Results

Total Ediacaran dataset

First, we analysed all the presence/absence data of organisms as a function of sites irrespective of their assemblage, in order to test whether the assemblage definitions represented distinct communities. The sites were ranked using reciprocal averaging ordination (y-axis of Fig 2), which provided a ranking that was consistent with previous work that grouped communities into the Avalon, White Sea, and Nama assemblages [7,8]. The coherence, turnover, and boundary clumping values were calculated, and the simulated mean was used to determine if that score was significantly different (Table 1). Table 1 gives the score, simulated mean, and significant level for each of the total set, Avalon, White Sea, and Nama assemblages, and environmental subsets of the assemblages. When analysing the total dataset, we found positive coherence, turnover, and boundary clumping, characteristic of a Clementsian-structured metacommunity (Fig 3A, Table 1). Site ordination scores were significantly associated with assemblage (H = 57.686, df = 2, p < 0.001, Appendix Table 4), indicating strong compositional difference among the assemblages. Site scores were also significantly correlated with depth (ρ = −0.427, p < 0.001, Table 2), suggesting that the structuring in the dataset may be due to the differences in depth between sites. However, since depth also significantly varies with assemblage (H = 53.987, df = 2, p < 0.001, Table 2), it is not possible to tell whether the structuring is due to depth or another factor that varies with assemblage.

Table 1

Metacommunity analyses.

Metacommunity values for coherence, turnover, and boundary clumping with interpretation of metacommunity structure within the EMS framework. Z is the Z-score, p is the p-value, and simMean is the simulated mean value of the metric.

	Coherence				Turnover				Boundary Clumping			Interpretation
Group	Coherence	p	simMean		Turnover	p	simMean		Moritsita’s Index	p
Total	2,230	<0.001	6,690	+	1,270,000	<0.001	1,010,000	+	6.390	<<0.001	+	Clementsian
Avalon	173	<0.001	309	+	5,120	0.007	7,650	-	1.880	<0.001	+	Clumped species loss (nested subsets)
Avalon (Margin slope)	167	<0.001	221	+	1,890	0.007	3,580	-	1.980	<0.001	+	Clumped species loss (nested subsets)
Avalon (Outer shelf)	10	0.510	11	+	0	0.119	21	-	1.330	0.156	+	No significant metacommunity structure
White Sea	695	<0.001	1,340	+	61,900	0.098	48,200	+	3.020	<0.001	+	Clementsian quasi structure
White Sea (Deep subtidal)	60	<0.001	217	+	2,780	0.230	2,030	+	3.450	<0.001	+	Clementsian quasi structure
White Sea (Middle shelf)	182	<0.001	344	+	5,300	0.110	3,580	+	4.700	<0.001	+	Clementsian quasi structure
White Sea (Outer shelf)	0	0.004	16	+	66	0.044	29	+	<0.001	0.227	-	Gleasonian
Nama	21	<0.001	133	+	688	0.435	640	+	2.350	<0.001	+	Clementsian quasi structure

EMS, Elements of Metacommunity Structure.

Fig 3

Metacommunity analyses and co-occurrence matrices for each assemblage.

(a) Metacommunity plot shows a summary of the metacommunity analyses. The z-score is a standardised way to indicate how much the observed means differ from the average across all communities in terms of standard deviations. Nested species loss is shown by an open circle; Clementsian by a closed circle; Gleasonian by a black ringed circle. The size of the circle represents the value of the boundary clumping score. The Avalonian outer shelf had a random structure and so is not shown. Co-occurrence matrices for (b) the Avalon metacommunity, (c) the White Sea metacommunity, and (d) the Nama metacommunities. Positive associations are blue; negative associations are yellow.

Table 2

Reciprocal averaging analyses.

Spearman rank correlations (Rs) (for continuous variables) and Kruskal–Wallis tests (for discrete variables) between site scores of each dataset obtained from reciprocal averaging and a variable, either assemblage or depth.

	Rs	Kruskal–Wallis	df	p-value
All assemblages site scores × assemblage	-	57.686	2	<0.001
All assemblages site scores × depth	−0.427	-	-	<0.001
Depth × assemblage	-	53.987	2	<0.001
Avalon site scores × depth	−0.360	-	-	0.051
White Sea site scores × depth	0.014	-	-	0.945
Nama site scores × depth	−0.728	-	-	0.007

To further investigate the nature of this structuring, we focused on the pairwise co-occurrence patterns, finding 10.3% were nonrandom: All but one positive associations resulted from taxa specific to the same assemblage (96.8% of positive associations), and all negative associations from taxa exclusively found in or heavily more abundant in different assemblages to each another (S2 Table) (100% of negative associations). The one exception was a positive association between Pteridinium and Rangea: These taxa are found roughly equally in Nama and White Sea sites with only a slight skew towards one or the other (S5 and S7 Tables). Therefore, our analyses are consistent with previous studies [8,9] in finding that Ediacaran taxa are highly segregated by assemblage, as well as confirming a role for depth (Table 2) in structuring the assemblages [7,8]. At this broad level of analysis, the strong assemblage signal, at least partially dependent on depth specialisations, obscures any other biotic or abiotic pattern.

Avalon metacommunity analyses

We then narrowed our level of analysis by focusing on each assemblage in turn. The Avalonian metacommunity displays significant positive coherence and boundary clumping, but significant negative turnover (Fig 3, Table 1), characteristic of a pattern of “nested clumped species loss” [37,45,46]. Site ordination scores were not significantly correlated with depth (Rs = −0.360, p = 0.051, Table 2) and so depth is only weakly associated with this metacommunity structuring. For Avalonian taxa, segregation was rare (Fig 3): Only one negative taxa association was found—between Charnia and Pectinifrons (Fig 3, S3 Table). Positive associations were found between Bradgatia and Charniodiscus, Fractofusus and Beothukis, Primocandelabrum, Charnia, Hadrynichorde, and Primocandelabrum (Fig 3, Table 2). The metacommunity structure was the same for the Avalonian margin slope metacommunity (the associations between Bradgatia and Charniodiscus and Fractofusus and Beothukis remained), but the outer shelf showed no significant structuring (Fig 2, S3 Fig, Table 1, S3 and S4 Tables). The predominance of positive over negative taxa co-occurrences is consistent with previous studies of detailed within-community spatial analyses of focal taxa, which found little evidence for lateral resource competition between Avalonian taxa [11,18]. The lack of depth and palaeoenvironmental correlation with metacommunity structure supports suggestions that Avalonian organisms have the widest niches and lowest provinciality among the Ediacaran biotas [7,12].

The Avalonian metacommunity displays a structure of “nested clumped species loss,” whereby taxa-poor communities form nested subsets of increasingly taxa-rich communities, with predictable patterns of taxa loss associated with variation in taxa characteristics [30]. Differences in Avalonian community composition have been suggested to represent different stages of community succession, based on community parameters, cluster analyses and MDS (Multidimensional Scaling) ordination [13]. Where multiple different stages of a community succession are analysed using EMS, the succession would result in the observed pattern of clumped taxa loss with early and late succession communities forming less diverse nested subsets of maximally diverse mid-succession communities. This metacommunity structure and proposed succession is consistent with Connell’s disturbance theory [13] whereby intermediate stages of a community are most diverse because they enable both early and late colonisers to coexist [47]. However, the lack of interspecific competition found by previous studies using spatial analyses within communities of early-stage communities [11,12,14] suggests that interactions other than competitive exclusion were influencing community development. Avalonian metacommunity structure had not previously been statistically compared to multiple different models (here 14 different models). Our results find that the Avalonian metacommunity exhibits clumped taxa loss, which supports Connell’s model. This model is further supported by the co-occurrence analyses, which found a negative association between late-succession and early-stage taxa (Charnia and Pectinifrons), and positive associations between late-stage (Primocandelabrum and Charnia), middle-stage (Bradgatia and Charniodiscus), and early-stage (Beothukis and Fractofusus) stage taxa (S2 Fig, S4 Table) [13].

The taxa associations of the Avalonian sites in the UK form a subset of the associations from the Canadian sites, suggesting structuring is not due to geographic or abiotic differences between the 2 areas (S3 Fig, S4 Table). Upon removing the UK sites from the analysis, positive associations between Fractofusus and Beothukis, and Hadrynichorde and Primocandelabrum remained significant (extended data Fig 4, S3 and S4 Tables). The association between Charniodiscus and Bradgatia was still present although not significant (p = 0.134) and the association between Charnia and Primocandelabrum was weakly significant (p = 0.054). The negative association between Charnia and Pectinifrons was also nearly significant (p = 0.089). There were also no new significant positive associations when studying only the Canadian sites and the associations thus form a perfect subset of the total dataset (S4 Fig, S6 Table). In Mistaken Point, Charnia is dominant in proposed late succession communities (on Lower Mistaken Point), whereas Pectinifrons is characteristic of early succession communities (on Shingle Head and D surface) [10,13]. These 2 taxa show a negative co-occurrence. Primocandelabrum, like Charnia, is characteristic of late- and middle-stage succession communities. Bradgatia and Charniodiscus are both overwhelmingly present in mid-succession communities. Beothukis and Fractofusus are most heavily found in early-mid-succession communities [13]. These pairs of taxa demonstrate positive co-occurrences. The Charnwood sites notably lack Fractofusus and Pectinifrons [48], both very characteristic of proposed early succession sites in Canada, but share many of the same taxa that are seen in proposed middle- and late-stage succession sites in Canada (Charnia, Primocandelabrum, Bradgatia, and Charniodiscus). There have been many fewer UK sites sampled than in Newfoundland, so it is plausible that the current communities do not reflect the full diversity of the area [49]. Thus, our results provide evidence that the UK and Canadian sites represent communities along similar community successions. Communities in the Flinders Range, South Australia (part of the White Sea assemblage) have been proposed to show evidence of primary succession [50], and so succession may be characteristic of many Ediacaran, as well as modern, communities. Therefore, observed differences in community composition between UK and Canadian Avalonian sites may reflect which stages in community development are preserved, and thus represent a biotic rather than a geographic signal.

Fig 4

Reconstructions of the Avalon, White Sea, and Nama metacommunities.

(A) A reconstruction of the Avalon assemblage showing the proposed stages of a community succession with the actual composition of several surfaces in boxes above. (a) Pectinifrons; (b) Beothukis; (c) Fractofusus; (d) Bradgatia; (e) Primocandelabrum; (f) Charnia; (g) Charniodiscus; (h) Culmofrons; (i) Trepassia. (B) A reconstruction of the White Sea assemblage showing some endemism of taxa to the Russian or Australian sites. (a) Charniodiscus; (b) Inaria; (c) Rangea; (d) Funisia; (e) Charnia; (f) Pteridinium; (g) Dickinsonia; (h) Rangea; (i) Tribrachidium; (j) Palaeopaschinus; (k) Coronacollina; (l) Albumares; (m) Kimberella; (n) Spriggina; (o) Parvancorina; (p) Rugoconites; (q) Eoandromeda; (r) Cyanorus; (s) Onega; (t) Armillifera; (u) Andiva; (v) Yorgia; (w) Temnoxa. (C) A reconstruction of the Nama assemblage showing the palaeoenvironmental separation of biomineralising and soft-bodied taxa across a depth profile. (a) Cloudina; (b) Namacalthus; (c) Ernietta; (d) Swartpuntia; (e) Nimbia; (f) Pteridinium; (g) Rangea. Taxa and environmental separation are not to scale. LMP, Lower Mistaken Point surface.

White Sea metacommunity analyses

The White Sea metacommunity displays significant positive coherence and boundary clumping and nonsignificant positive turnover, characteristic of a “Clementsian quasi structure” (S1 Table) [30] with no correlation between site ordination scores and depth (Rs = 0.014, p = 0.945, Fig 2; Tables 1 and 2), and so depth alone is unlikely to be responsible for this metacommunity structure (Table 2). Metacommunity structure within the White Sea assemblage differed with palaeoenvironment: The deep subtidal and middle shelf metacommunities displayed “Clementsian quasi structures,” while the outer shelf metacommunity was characterised by a Gleasonian structure, with significant positive coherence and turnover and nonsignificant boundary clumping (Fig 3, S4 Fig, Table 1). Co-occurrence analyses found that 11 of the 32 positive associations found for the whole assemblage were preserved when focusing on the palaeoenvironmental subsets (Fig 3, S4 Fig, S6 and S7 Tables). A positive association between Parvancorina and Tribrachidium was the only association to appear in both environmental subdivisions despite almost all the White Sea taxa being present in both the middle shelf and deep subtidal environments. In deep subtidal facies, there is a notable negative association between Beltanelliformis and Kimberella (S5 Fig, S5–S7 Tables).

Several taxa seemed to have unique taxa associations in each subdivision despite very similar community composition. (S4 Fig, S6 and S7 Tables), which naively we would expect to lead to the same taxa associations. The 11/32 positive associations that differed between the middle shelf and deep subtidal environments show that the community associations are nonconsistent between the subsets and the assemblage as a whole. The underlying processes that contribute to these differences in both intertaxa interactions and environmental factors could be due to organism behavioural plasticity, leading to different behaviours in different environments. Alternatively, differences in taxa associations for a given taxon may reflect the inclusion of several taxa with different environmental preferences and behaviours within one taxonomic group (e.g., Dickinsonia). However, as most Ediacaran taxonomic groupings are monotypic the differences in taxa associations in different environments are more likely due to plastic responses to, e.g., variation in resource limitation or the presence of different competitors or ecosystem engineers.

The majority of taxa pairwise associations found in the Russian White Sea are also found in the pooled White Sea metacommunity (24/37). However, 13 of the pooled associations are present when only analysing Russian localities, suggesting that some of the structure in the dataset may be due to geography (S5 and S8 Tables). There is marked geographic variation in community composition between the Russian and Australian White Sea localities [24], so the nonshared associations may reflect a greater endemism within the White Sea assemblage compared to the Avalonian assemblage, where the UK sites formed a perfect subset of the Canadian sites.

The only evidence of a putative consumer–resource interaction was found in the White Sea metacommunity, which is consistent with the idea that grazing and motility evolved as part of the “second-wave radiation,” where the first-wave radiation was the “Avalon Explosion” [5,51,52]. In White Sea deep subtidal facies, there was a negative association between Beltanelliformis and Kimberella. This result may be a consequence of herbivory as Kimberella has been reconstructed as a mobile grazer [53,54] and Beltanelliformis as large colonies of cyanobacteria [55]. It has also previously been noted that remains of B. brunsae sometimes co-occur with the feeding traces of Kimberella (Kimberichnus teruzzii) [56,57].

Nama metacommunity analyses

The Nama metacommunity has significant positive coherence and boundary clumping and nonsignificant positive turnover and so displays a “Clementsian quasi structure,” the same metacommunity structure as the White Sea assemblage (Table 1) [30]. Unlike the Avalon and White Sea sites, ordination scores were significantly correlated with depth (Rs = −0.728, p = 0.007, Table 2), and so the Clementsian structuring occurs along a depth gradient. Pairwise taxa co-occurrences revealed significant negative associations between biomineralisers (Cloudina) and soft-bodied taxa (Pteridinium and Rangea) and a significant positive association between 2 soft-bodied taxa (Fig 3, S10 Table). There were more negative than positive associations (Fig 3, S10 Table). These results statistically confirm previous observations of separation between biomineralisers (e.g., Cloudina) and soft-bodied organisms (such as Pteridinium and Rangea) [58].

These patterns of segregation are unlikely to be purely facies-based control for our data set, which is purely a result of the chemical and physical properties of the rock that the fossils are preserved in. In the Nama assemblage, there are both deep and shallow water carbonate facies. The deep water Nama facies (Dengying Fm) has a more similar community composition to the other deep water Nama sites than to the carbonate sites and thus help to strengthen the pattern of taxa segregation by depth as opposed to counteract it (which is what we would expect if there was purely facies-based control of taxa separation). If the habitat specialisation was a reflection on biomineralisation alone, that we would expect to see the biomineralisers behave in broadly similar patterns, and the soft-bodied taxa to also behave similarly to each other. Of the 7 taxa that were sufficiently abundant to be included in these analyses, one was a putative microbial colony (Nimbia [59]), 2 were biomineralisers (Cloudina and Namacalathus), and the remaining 4 were soft-bodied taxa (Rangea, Pteridinium, Ernietta, and Swartpuntia). Nimbia did not show any significant associations with the other taxa, although (like Namacalathus) it was only present in 2 sites, this lack of associations may be due in part to small sample sizes. The 2 biomineralisers behaved in different ways—while Cloudina showed negative associations with the soft-bodied Rangea and Pteridium. Namacalathus did not show any significant associations with any taxa but could be biased by the number of sites in which it is present. The soft-bodied taxa also did not behave in a uniform way, with Ernietta and Swartpuntia displaying no significant associations with either other soft-bodied taxa nor biomineralisers while Rangea and Pteridium showed significant positive association with each other and negative associations with Cloudina. As such, there are no consistent patterns of biomineralisers nor soft-bodied taxa that explain the patterns found within our data.

As such, the signal in our data cannot be attributed solely to carbonate/siliclastic nor biomineralisers/soft-bodied taxa differences, and so is most likely due to habitat preferences as community composition was found to vary significantly with depth. Cloudina is found exclusively in shallow limestone and shallow siliciclastic shoreface facies, whereas soft-bodied Nama taxa are found in both deeper shoreface and deep subtidal settings (Fig 4).

Effect of sampling biases

The Nama assemblage has notably less localities [9] than either the Avalon [29] or the White Sea [28], which could suggest that differences in the Nama are merely an artefact of sampling. Therefore, it is important to understand how these sampling differences could affect the EMS analyses and the co-occurrence analyses. We assessed these biases in 2 ways: (1) by comparing results of environmental subsets of the Avalon and White Sea, which are similar in size to the Nama assemblage; and (2) by simulating Avalon and White Sea data by subsampling the larger datasets to that of Nama—9 localities then testing for significant nonrandom co-occurrences and for a correlation between site score and depth.

First, in terms of co-occurrence, for the Avalon subsets, the margin slope [23] and outer shelf [6] have 4.5% and 0% significant nonrandom co-occurrence, and for the White Sea, the deep subtidal has 4.1% significant co-occurrences and middle shelf has 7.9%. These values are much smaller than that of the Nama at 16.7%. Furthermore, they show an increase from the Avalon to the White Sea, thus confirming the overarching pattern of increasing co-occurrences found in the full sets. In terms of metacommunity structure, the Avalon environmental subsets have the same metacommunity structure (nested clumped species loss) as the whole Avalon, with negative turnover and small coherence factors. Similarly, the White Sea environmental subsets both have the same quasi-Clementsian structure as the whole White Sea assemblage. Thus, the changes of metacommunity structure from the Avalon to the White Sea are maintained within the environmental subsets with sample sizes similar to those available for Nama.

Second, we performed randomised tests for the co-occurrence and depth analyses for the Avalon and White Sea data, subsampling the datasets 1,000 times each from 29 (Avalon) and 28 (White Sea) to 9 (Nama). Avalon had significantly less nonrandom co-occurrences than the While Sea and Nama (p = 0.016; p = 0.016), in contrast to the White Sea, which showed no significant difference in numbers of co-occurrences (p = 0.158). In order to test for the significance of the depth correlation we performed Spearman test each of the 1,000 subsampled data. Only a small number of subsamples showed a significant correlation for depth (24 out 1,000 for Avalon; and 17 out of 1,000 for the White Sea). Therefore, we are confident that our results showing an increase in co-occurrence between the Avalon and Nama and an increase in depth structure between the White Sea and Nama are not artefacts of different sampling, but robust signals.

Discussion

The total set of Ediacaran data exhibits strong metacommunity structure, consistent with previous analyses that resolve multiple assemblages [7-9]. Both the total dataset and the individual assemblages have relatively low numbers of nonrandom co-occurrences (9.8% to 16.7%) compared to many extant analyses (e.g., 35% to 63% [60,61]) as well as terrestrial fossil communities (such as averaging 64% aggregated pairs from the Carboniferous to the Holocene, and 37% from the Holocene to the present [62], although percentages of nonrandom co-occurrences are similar to at least some extant benthic communities (16.3%) [63]. Previous spatial analyses of Avalonian communities have revealed limited interspecific interactions [11,14] and limited environmental associations between taxa within communities [12], so the large number of nonsignificant correlations within the Avalon assemblage are consistent with previous work. Given that the oldest assemblage is dominated by non-co-occurrences, it follows that the subsequent development of Ediacaran metacommunities would have to build from this point and so are not immediately comparable to extant communities with longer evolutionary histories. There is strong evidence for the second-wave diversification reflected in our results in an increase of nonrandom taxa associations from the Avalonian biota (9.8%) to the White Sea biota (16.1%, p = 0.016). Further evidence for this increased ecological complexity (i.e., greater interactions and associations between taxa) is provided by the fact that the Avalonian assemblage had minimal environmental structuring, while we detected both Gleasonian and quasi-Clementsian metacommunity structure depending on the palaeoenvironments for the White Sea assemblage. While there is not a significant correlation of broad-scale palaeoenvironment with metacommunity structure, these structures reflect a significant influence of a fine-scale environmental gradient. Gleasonian structuring reflects an individualistic response to the inferred environmental gradient, suggesting a lack of within-community associations for this outer shelf metacommunity. In contrast, quasi-Clementsian structure corresponds to a community-wide response to the environmental gradients, so reflects within-community specialisations in the White Sea assemblage. These specialisations are reflected in behavioural flexibility, with organisms exhibiting different taxa associations in different environments across a wide range of depths.

The White Sea and Nama assemblages have the same type of metacommunity structure as shown through EMS, with both assemblages showing a quasi-Clementsian structure (Fig 3, Table 1). There is a small increase in nonrandom associations in the Nama biota (16.7% from 16.1% in the White Sea), which, like the EMS analyses, shows at least a maintenance, if not slight increase, in metacommunity structure. There is a significant increase in ecosystem structuring between the White Sea and Nama assemblages when site rank within each of the assemblages are compared to depth (Table 2). Neither the Avalon nor White Sea show such significant correlation with depth, in sharp contrast to the Nama, where site composition is significantly correlated with depth (Table 2). Taken together, these 3 analyses show that compared to the White Sea, the Nama assemblage has an increased taxa segregation coupled to a strong palaeoenvironmental correlation and therefore narrowed environmental tolerances, showing a further decrease of niche breadth. Thus, we have shown that the increase in complexity of the taxon-specific ecological strategies utilised throughout the Ediacaran is mirrored in the complexity of the community associations.

A White Sea–Nama catastrophic environmental extinction is in consistent with our results for 3 reasons. Firstly, a catastrophic mass extinction implies that surviving taxa within the Nama assemblage are more likely to be generalists [8,9,64], contrary to our results. We have shown an increased influence of paleoenvironment and niche specialisation with the Nama metacommunity showing significant correlation between community composition and depth, in contrast to the White Sea and Avalon metacommunities. Second, the Nama metacommunity exhibits a nonsignificant but positive turnover, indicating more turnover along the gradient (more niche differentiation). If the Nama assemblage metacommunity structure was due to underlying extinction/colonisation dynamics, we expect to see an increase in nestedness [40], as indicated by negative turnover, contrary to our positive Nama turnover (Table 1). Thirdly, this increase in turnover suggests not only higher ecosystem complexity but also increased taxa specialisation and narrower niche breadth coupled with an increase in within-community structuring between the White Sea and Nama assemblages, with a slight increase in nonrandom taxa associations (16.1% to 16.7%). A decrease in taxonomic and morphological diversity in the Nama [23] may reflect that, within this assemblage, Ediacaran organisms show significant palaeoenvironmental preferences, and thus reduced environmental tolerances, resulting in multiple different types of mutually exclusive communities, each of which exhibits a simple structure within its narrow niche [23]. An increase in within-community structure in the form of ecosystem engineering [65] and reef complexity [20] provides supporting evidence that despite a decrease in taxonomic diversity, the Nama assemblage represents an ecological development from the White Sea assemblage, not a recovery from a catastrophic extinction event. Our results are further supported by birth–death models of stem and crown group diversification, which predict Ediacaran-like diversification patterns for bilaterians and produce patterns that can be easily mistaken for mass extinctions [66].

Our results show that these Ediacaran organisms underwent the niche contraction and specialisation that is traditionally associated with Cambrian diversification [6,67]. Therefore, we find that the eco-evolutionary dynamics of metazoan diversification known from the Cambrian started earlier in the Ediacaran with the Avalon assemblage and increased in complexity towards the Phanerozoic as new anatomical innovations appeared, culminating in the “Cambrian Explosion.”

Materials and methods

Materials

The data used in this study is a binary presence/absence matrix for 86 Ediacaran localities and 124 taxa. The data is taken from [9] with more conservative classifications of assemblages for several sites (cf., [8]): SB-Nor2 and SB-So1, both Sewki Brook sites, are classed as indeterminate in our analysis and [8] but as Avalon and Nama, respectively, in [9]. Two Chinese sites classed as Nama by Muscente and colleagues are classed as indeterminate here (Gaojiashan and Lijiagou) [8,24]. The data contains information on the palaeoenvironment, depth index (from 1 to 11), lithology, and assemblage of each locality as well as a time index (from 1 to 3). The full classification of sites in the dataset can be seen in S1 Data and the palaeoenvironmental and assemblage classifications in Fig 1. This data is appropriate for applying modern statistical ecological methods because the organisms were mostly sessile and benthic and preserved in such a way that they are interpreted as in situ life assemblages with minimal transportation after death or time-averaging [13,26-28].

The Avalon, White Sea, and Nama assemblages are represented by 30, 29, and 12 sites, respectively, with 11 undetermined sites and 4 that belong to the Doushantuo assemblage. The Doushantuo assemblage was excluded from the assemblage-level analyses because there are only 4 sites, which is insufficient to run these analyses.

Secondly, we used Spearman rank correlations to test whether within-assemblage community composition is correlated with depth. The ordering of the sites was given by the ordination output from the EMS analyses (Fig 2), which is produced by reciprocal averaging, a type of correspondence analysis that ordinates the sites based on their species composition [31]. This ordering provides the first-axis ranking of the sites that were then used to test whether there was a significant association with depth.

Methods

The R package Metacom was used for the EMS analyses [37]. The first step of the EMS analyses was to use reciprocal averaging to ordinate the sites based on their species composition ([31]; Fig 2). Metacommunity structure is then quantified based on this ordering via the calculation of the 3 metrics related to metacommunity structure: coherence, turnover, and boundary clumping [37]. A 3-tiered analysis based on these metrics enabled the placement of each studied metacommunity into one of 14 idealised metacommunity structures following the EMS approach [29,30]. In the context of EMS, statistical significance is calculated using a z-test (i.e., calculating the z-score) of the observed absences to embedded absences in randomised null matrices [31]. For each subsection of data (e.g., each assemblage), the z-score is calculated relative to that subset only, rather than the total set of data. Significant coherence is a prerequisite for further analysis of metacommunity structure, and so this was the first metric calculated. Calculation of turnover was then used to distinguish whether the metacommunity formed a nested structure. Calculation of boundary clumping allowed the determination of whether the taxa range boundaries were clumped, dispersed, or not significantly correlated with each other along the environmental gradient. Fig 1 gives examples of taxa abundance distributions that give rise to the idealised metacommunity structures. Sites scores were given according to the ranking of the sites in the first-degree ordination of their taxa composition. These were used to investigate the importance of depth (as indicated by palaeoenvironment) and assemblage variables in structuring taxa distributions via a nonparametric Spearman correlation or Kruskal–Wallis test. The metacommunity analyses were performed on the entire dataset, for each assemblage individually, and for palaeoenvironmental and geographic subsets within each assemblage where there were enough localities for valid analyses.

Co-occurrence analysis

The R package co-occur was used to calculate the observed and expected frequency of co-occurrence between pairs of taxa to determine significant positive or negative associations [42]. Taxa, which only occurred in 1 site, were removed from the analysis because such singletons have been shown to disproportionally influence co-occurrence analyses [68,69]. Co-occurrence analysis was done for the entire dataset, for each assemblage individually, and for palaeoenvironmental and geographic subsets within each assemblage where there were enough sites [42].

Locality map showing Ediacaran sites.

The names, assemblages, and various characteristics of each of the localities can be found in Fig 1 and in S1 Data. The map is from generated in R using the ggplot2 and sf packages in R, using an OpenStreetMap basemap (71).

(EPS)

Click here for additional data file.

Co-occurrence matrix for the total dataset.

Co-occurrence matrix for species showing significant associations in the whole dataset. Positive associations are blue; negative associations are yellow.

(EPS)

Click here for additional data file.

Co-occurrence matrices for the Avalonian dataset and subsets.

Co-occurrence matrices for species showing significant associations in (a) the Canadian Avalonian metacommunity, (b) the Avalonian margin slope metacommunity, and (c) the Avalonian outer shelf metacommunity. Positive associations are blue; negative associations are yellow. The Avalonian outer slope metacommunity had no significant associations.

(EPS)

Click here for additional data file.

Co-occurrence matrices for the White Sea dataset and subsets.

Co-occurrence matrices for species showing significant associations in (a) the Russian White Sea metacommunity, (b) the White Sea middle shelf metacommunity, and (c) the White Sea deep subtidal metacommunity. Positive associations are blue; negative associations are yellow.

(EPS)

Click here for additional data file.

Summary table of how metacommunity properties are expressed in terms of the EMS metrics.

(DOCX)

Click here for additional data file.

Co-occurrence analysis for the total dataset showing only significant associations.

Sp1_inc is the number of sites that have taxa 1. Obc_cooccur is the observed number of sites with both species. Prob_cooccur is the probability both species occur at a site. Exp_cooccur is the expected number of sites having both taxa. P_Lt is the probability that the 2 taxa would co-occur at a frequency less than observed, and P_gt is the probability that the 2 taxa would co-occur at a frequency greater than observed. Difference is the difference between observed and expected probabilities, where difference > 0.95 the association is considered significant.

(DOCX)

Click here for additional data file.

Co-occurrence analysis for the Avalonian dataset showing only significant associations.

Sp1_inc is the number of sites that have taxa 1. Obc_cooccur is the observed number of sites with both species. Prob_cooccur is the probability both species occur at a site. Exp_cooccur is the expected number of sites having both taxa. P_Lt is the probability that the 2 taxa would co-occur at a frequency less than observed, and P_gt is the probability that the 2 taxa would co-occur at a frequency greater than observed. Difference is the difference between observed and expected probabilities, where difference > 0.95 the association is considered significant.

(DOCX)

Click here for additional data file.

Co-occurrence analysis for the Avalonian Canada dataset showing only significant associations.

Sp1_inc is the number of sites that have taxa 1. Obc_cooccur is the observed number of sites with both species. Prob_cooccur is the probability both species occur at a site. Exp_cooccur is the expected number of sites having both taxa. P_Lt is the probability that the 2 taxa would co-occur at a frequency less than observed, and P_gt is the probability that the 2 taxa would co-occur at a frequency greater than observed. Difference is the difference between observed and expected probabilities, where difference > 0.95 the association is considered significant.

(DOCX)

Click here for additional data file.

Co-occurrence analysis for the Avalonian margin slope dataset showing only significant associations.

Sp1_inc is the number of sites that have taxa 1. Obc_cooccur is the observed number of sites with both species. Prob_cooccur is the probability both species occur at a site. Exp_cooccur is the expected number of sites having both taxa. P_Lt is the probability that the 2 taxa would co-occur at a frequency less than observed, and P_gt is the probability that the 2 taxa would co-occur at a frequency greater than observed. Difference is the difference between observed and expected probabilities, where difference > 0.95 the association is considered significant.

(DOCX)

Click here for additional data file.

Co-occurrence analysis for the White Sea dataset showing only significant associations.

Sp1_inc is the number of sites that have taxa 1. Obc_cooccur is the observed number of sites with both species. Prob_cooccur is the probability both species occur at a site. Exp_cooccur is the expected number of sites having both taxa. P_Lt is the probability that the 2 taxa would co-occur at a frequency less than observed, and P_gt is the probability that the 2 taxa would co-occur at a frequency greater than observed. Difference is the difference between observed and expected probabilities, where difference > 0.95 the association is considered significant.

(DOCX)

Click here for additional data file.

Co-occurrence analysis for the White Sea middle shelf dataset showing only significant associations.

Sp1_inc is the number of sites that have taxa 1. Obc_cooccur is the observed number of sites with both species. Prob_cooccur is the probability both species occur at a site. Exp_cooccur is the expected number of sites having both taxa. P_Lt is the probability that the 2 taxa would co-occur at a frequency less than observed, and P_gt is the probability that the 2 taxa would co-occur at a frequency greater than observed. Difference is the difference between observed and expected probabilities, where difference > 0.95 the association is considered significant.

(DOCX)

Click here for additional data file.

Co-occurrence analysis for the White Sea deep subtidal dataset showing only significant associations.

Sp1_inc is the number of sites that have taxa 1. Obc_cooccur is the observed number of sites with both species. Prob_cooccur is the probability both species occur at a site. Exp_cooccur is the expected number of sites having both taxa. P_Lt is the probability that the 2 taxa would co-occur at a frequency less than observed, and P_gt is the probability that the 2 taxa would co-occur at a frequency greater than observed. Difference is the difference between observed and expected probabilities, where difference > 0.95 the association is considered significant.

(DOCX)

Click here for additional data file.

Co-occurrence analysis for the White Sea Russian dataset showing only significant associations.

(DOCX)

Click here for additional data file.

Co-occurrence analysis for the Nama dataset showing only significant associations.

Sp1_inc is the number of sites that have taxa 1. Obc_cooccur is the observed number of sites with both species. Prob_cooccur is the probability both species occur at a site. Exp_cooccur is the expected number of sites having both taxa. P_Lt is the probability that the 2 taxa would co-occur at a frequency less than observed, and P_gt is the probability that the 2 taxa would co-occur at a frequency greater than observed. Difference is the difference between observed and expected probabilities, where difference > 0.95 the association is considered significant.

(DOCX)

Click here for additional data file.

Supplementary data.

(CSV)

Click here for additional data file.

Supplementary code.

(TXT)

Click here for additional data file.

12 May 2021

Dear Dr Mitchell,

Thank you for submitting your manuscript entitled "Metacommunity analyses show increase in ecological specialisation throughout the Ediacaran" for consideration as a Research Article by PLOS Biology.

Your manuscript has now been evaluated by the PLOS Biology editorial staff, as well as by an academic editor with relevant expertise, and I'm writing to let you know that we would like to send your submission out for external peer review.

However, before we can send your manuscript to reviewers, we need you to complete your submission by providing the metadata that is required for full assessment. To this end, please login to Editorial Manager where you will find the paper in the 'Submissions Needing Revisions' folder on your homepage. Please click 'Revise Submission' from the Action Links and complete all additional questions in the submission questionnaire.

Please re-submit your manuscript within two working days, i.e. by May 14 2021 11:59PM.

Login to Editorial Manager here: https://www.editorialmanager.com/pbiology

During resubmission, you will be invited to opt-in to posting your pre-review manuscript as a bioRxiv preprint. Visit http://journals.plos.org/plosbiology/s/preprints for full details. If you consent to posting your current manuscript as a preprint, please upload a single Preprint PDF when you re-submit.

Once your full submission is complete, your paper will undergo a series of checks in preparation for peer review. Once your manuscript has passed all checks it will be sent out for review.

Given the disruptions resulting from the ongoing COVID-19 pandemic, please expect delays in the editorial process. We apologise in advance for any inconvenience caused and will do our best to minimize impact as far as possible.

Feel free to email us at plosbiology@plos.org if you have any queries relating to your submission.

Kind regards,

Roli Roberts

Roland Roberts

Senior Editor

PLOS Biology

rroberts@plos.org

16 Jul 2021

Dear Dr Mitchell,

Thank you very much for submitting your manuscript "Metacommunity analyses show increase in ecological specialisation throughout the Ediacaran" for consideration as a Research Article at PLOS Biology. Your manuscript has been evaluated by the PLOS Biology editors, an Academic Editor with relevant expertise, and by three independent reviewers. Please accept my apologies for the delay in delivering a decision; we've had the reviews for a while but had difficulties contacting the Academic Editor to discuss the way forward.

IMPORTANT: You'll see that the reviewers are all broadly positive about your study. However, they clearly feel that the manuscript needs a lot of work in two main areas: first improving the clarity, flow and structure of the content, to achieve a fluid and clear presentation of the results. The Figures will need to flow in parallel with the main text, which will require revising these and improving the presentation, paying additional attention to the legends (which should be more self-contained).

The referees (especially rev #1 and #2) also raise some technical issues that will need to be carefully addressed: rev #1 on modularity, and rev #2 regarding recent issues and criticism of inferences of interactions from simple co-occurrence data.

In light of the reviews (below), we will not be able to accept the current version of the manuscript, but we would welcome re-submission of a much-revised version that takes into account the reviewers' comments. We cannot make any decision about publication until we have seen the revised manuscript and your response to the reviewers' comments. Your revised manuscript is also likely to be sent for further evaluation by the reviewers.

We expect to receive your revised manuscript within 3 months.

Please email us (plosbiology@plos.org) if you have any questions or concerns, or would like to request an extension. At this stage, your manuscript remains formally under active consideration at our journal; please notify us by email if you do not intend to submit a revision so that we may end consideration of the manuscript at PLOS Biology.

**IMPORTANT - SUBMITTING YOUR REVISION**

Your revisions should address the specific points made by each reviewer. Please submit the following files along with your revised manuscript:

1. A 'Response to Reviewers' file - this should detail your responses to the editorial requests, present a point-by-point response to all of the reviewers' comments, and indicate the changes made to the manuscript.

*NOTE: In your point by point response to the reviewers, please provide the full context of each review. Do not selectively quote paragraphs or sentences to reply to. The entire set of reviewer comments should be present in full and each specific point should be responded to individually, point by point.

You should also cite any additional relevant literature that has been published since the original submission and mention any additional citations in your response.

2. In addition to a clean copy of the manuscript, please also upload a 'track-changes' version of your manuscript that specifies the edits made. This should be uploaded as a "Related" file type.

*Re-submission Checklist*

When you are ready to resubmit your revised manuscript, please refer to this re-submission checklist: https://plos.io/Biology_Checklist

To submit a revised version of your manuscript, please go to https://www.editorialmanager.com/pbiology/ and log in as an Author. Click the link labelled 'Submissions Needing Revision' where you will find your submission record.

Please make sure to read the following important policies and guidelines while preparing your revision:

*Published Peer Review*

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out. Please see here for more details:

https://blogs.plos.org/plos/2019/05/plos-journals-now-open-for-published-peer-review/

*PLOS Data Policy*

Please note that as a condition of publication PLOS' data policy (http://journals.plos.org/plosbiology/s/data-availability) requires that you make available all data used to draw the conclusions arrived at in your manuscript. If you have not already done so, you must include any data used in your manuscript either in appropriate repositories, within the body of the manuscript, or as supporting information (N.B. this includes any numerical values that were used to generate graphs, histograms etc.). For an example see here: http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1001908#s5

*Blot and Gel Data Policy*

We require the original, uncropped and minimally adjusted images supporting all blot and gel results reported in an article's figures or Supporting Information files. We will require these files before a manuscript can be accepted so please prepare them now, if you have not already uploaded them. Please carefully read our guidelines for how to prepare and upload this data: https://journals.plos.org/plosbiology/s/figures#loc-blot-and-gel-reporting-requirements

*Protocols deposition*

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Thank you again for your submission to our journal. We hope that our editorial process has been constructive thus far, and we welcome your feedback at any time. Please don't hesitate to contact us if you have any questions or comments.

Sincerely,

Roli Roberts

Roland Roberts

Senior Editor

PLOS Biology

rroberts@plos.org

*****************************************************

REVIEWERS' COMMENTS:

Reviewer #1:

The authors present here a series of analyses of presence/absence data for 124 taxa across 86 Ediacaran localities to examine whether a large drop in diversity in the late Ediacaran was due to an external disturbance (mass extinction) or the results of eco/evolutionary forces such as niche specialization and competition restructuring the communities. The authors claim that their analyses point to the latter, and while I don't have strong reason to doubt this claim, the conclusions that they draw from their analyses are very hard to follow. In large part this is due to the fact that they a) don't adequately describe the methods they use, and b) don't clearly point the reader to how changes in one metric or another leads to different interpretations and why this should be the case. In part this can be helped by porting some of the basic explanations of what is being done and why to the fore of the manuscript, with details in the methods of course, but even the methods are lacking in terms of specifics. While I understand that the analyses were created and reported elsewhere, I think it is important for the basics to be reported here - especially for the 3 metrics used to parse potential community organizations that they are examining. Apart from the minimal methods, I found the objectives of the paper to be hard to follow. Really what the manuscript needs most is motivation. Why is it important to examine potential changes in community structure, and why are the methods employed the appropriate ones to use in this case? For example - the authors state on L58 "we will investigate the structuring mechanisms within these assemblages". While structure is clearly assessed, it is much less clear how the authors extract *mechanism* from differences in measured structural attributes. While the authors may be quite justified in their interpretation of mechanism, it needs to be clearly stated, which it currently is not. As it stands there are so many unclear aspects of the manuscript, I find it impossible to judge in terms of its technical merits.

My interpretation then is that this paper has a lot of potential - it is an interesting look at a very old community (well, communities). The conclusions that the authors describe are also very interesting as they suggest that a previously interpreted mass extinction may in fact be due to expected changes in the eco/evolutionary trajectories of these communities without such a dramatic externality such as a mass extinction to bring about community change. The pieces are all in place, but the authors need to spend a little more effort to connect them, in my humble opinion. I'd be excited to review an revision that incorporates some of these - mainly textual rather than technical - changes. I was also a bit surprised at the number of grammatical errors and incomplete sentences throughout the manuscript, as if things were quickly shuffled around prior to submission. I'd advise the authors to make make sure that all of their sentences follow grammatical rules during revision and that sentence structure is complete. On a more minor note, ensure that American/British spellings are corrected - both 'specialisation' and 'specialization' are used. Not sure which Plos prefers.

L19 - The authors might consider making it more clear that this is a reinterpretation of the 'diversity drop' described in the first sentence of the abstract, rather than an independent (and not necessarily mutually exclusive) finding.

L21 - Moreover, I think that it is more correct to say that this 'diversity drop' can be explained by an alternative hypothesis (structuring and resulting competitive exclusion, etc)... it is not certain it is one or the other (mass extinction), correct?

L51-53 - I'm not sure what 'using modularity to compare' means... modularity is a network property, but it may be more correct to say that they found differences in the compartmentalization of XXX in the networks rather than to say they use the metric to compare (as it's not a robust statistic for directly comparing networks on its own). Also it is a bit jargony the way it is used here, so some clarification would help. The term 'cluster' used on L53 is also not clear in the way it is used, as it is stated it is a prior analysis, but does this mean it is the same analysis as the one that employs modularity? Overall, the description of the prior work - both the means and the primary conclusions - need to be clarified. It is also important to specify what network is being used. I assume co-occurrence?

L59 - A detail, but 'using a matrix' is less clear than the fact that the authors compiled presence/absence for 86 localities, 124 taxa (wow!)... Whether or not you put this information into a matrix format is less material... If this matrix format is important for the EMS framework, I would suppose it's organization is also important, so should be described a bit more clearly (there are many ways this information can be put into a matrix).

L67-68 - these metrics need to be defined. They can mean different things in different contexts, and it should be assumed that the reader is not familiar with the EMS framework.

L69 - what does 'fits within the EMS framework' mean? Is there a statistical measure whereby these metrics tell us whether the inputted species across localities are functioning as a spatial meta community? It would be more helpful - rather than to refer to the 'EMS framework' - to tell the reader what the methods employed are telling us ecologically. In other words, I care less about the framework than I care about what the framework tells us regarding the ecology, so the ecology should be put center stage, with the framework being a means to an end.

Figure 1 - Cool figure, and I think I get the descriptions of the outcomes, but the mechanisms by which these outcomes supposedly derive are completely mysterious. Also - I can't seem to find what NS means… Non-structured? Another very odd thing about this figure is that - for example - coherence is split from left to right as (-) (+) (0)… the other metrics are presented in various order from left to right. Because these metrics are continuous variables, this swapping around in various ways is confusing. If there is a way to order the resulting community structures in a way that makes sense with respect to the continuous nature of the variables?

L70-72 - Can you describe what you mean by random or non-random? There are many ways by which species could be randomly distributed across sites. In the case of this test, what does random mean. More importantly, what does random and non-random imply eco/evolutionarily. The reader needs to be pointed to what these different conditions mean in terms of our interpretation of the community.

L72 - what is reciprocal averaging? Averaging of what?

L67-74 - Both in the abstract and later in the paper, a central point appears to be posing eco/evo restructuring of the meta community as an alternative hypothesis for late Ediacaran diversity declines. If that is a central aim, I would think it would be mentioned here? Overall I think the aims need to be motivated by how they will be filling in knowledge gaps… i.e. you are listing what you intend to do, but not 'why'.

Regarding this and prior comments on clarity, I understand that the details are in the Materials and Methods, but the Intro/Results/Discussion shouldn't be totally mysterious and jargony. The material needs to be introduced and described in enough detail for the reader to understand why things are being done and what this will mean, even if they will later need to dig into the methods to get the details. So far, I am helplessly adrift as I read through the intro, and I don't know why things are being done and for what reason. Apart from an illusion in the abstract, I'm not even clear that you are testing an alternative hypothesis for the decline in late Ediacaran diversity. The Intro really needs to be re-envisioned to emphasize the aims of the paper, and the motivation for applying the analyses that are described... and what this can perspective can shed light on. [Circling back after reading the materials and methods, they give me a sense of what the measures are measuring, but I don't have a sense of the details. Things like z-scores, which are reported, are not mentioned].

L93-96 - This is a cool finding!

Table 1 - It would be very helpful to orient the readers - in a very basic way - how changes in the three metrics are used to identify the different listed interpretations... i.e. High coherence and low turnover and clumping lead to XXX communities... Low coherence and high turnover lead to YYY communities, etc. Also need to define in Layman's terms what coherence/turnover/clumping means in this context. Not dissimilar from what is written on L155-157, but as a roadmap for the different potential end-conditions in Figure 1. [Circling back, this is done a bit in materials and methods - some of this needs to be placed in the fore, so that the reader can follow what changes in these metrics represent in terms of community structure]

Figure 3 - I can find the (general) definition for boundary clumping (though I think all metrics: coherence, turnover, boundary clumping should be quantitatively defined here - otherwise I have no way of understanding how or why different sites have the values they do), but I have no idea what a 'clump' is, as 1, 3, 6 clumps are reported in Figure 3. Even more problematically, nowhere in the manuscript can I find reference to the z-value displayed in the figure. Because only the raw values of coherence, turnover, and boundary clumping are discussed… with important meanings for (+,-,0), it would be helpful to report why it is preferable to look at the z-values (presumably z-scores?). For the z-scores, what is serving as the mean and sd used to compute values for individual sites? In other words, is the Avalon z score: (Avalon x - Avalon mean) / Avalon sd? Or is it (Avalon x - Total mean)/(Total sd)?

L142 - Wait... what are the 12 models? Either I missed it or it is not mentioned prior to this reference, which is confusing and a bit disconcerting.

L168-171 - Not to be a stats snob, but is this significantly different in a statistical sense? Not everything needs a p-value, but you might consider a different adjective if you use significantly different in a statistical sense elsewhere.

L171 - I think this is a stretch... you don't have interaction information, just co-occurrence information which may or may not (often not) correlate with interactions. I think what is needed here is an argument that the occurrence in different communities implies a change in interactions between species. I would think you could only truly support this if species A is in a completely unique environment where there are no other co-occurring species between the communities being compared. Then species A must be interacting with completely different species in each site. If there are even a small number of co-occurring species in 2 communities with species A, species A might not be changing its behavior - it may just have stronger interactions with the subset of species that co-occur in both sites. So some consideration of the co-occurrence vs. interaction concepts is important here.

L197 - very interesting result!

Figure 4 - this is such a cool figure

L222 - A period missing between 'depth' and 'Claudina'?

L240 - Seems like not a full sentence - something appears to be missing as the next sentence discusses a 'final radiation' but no context wrt earlier radiations.

L247 - Is complexity here being used as a synonym for high community heterogeneity between environments?

L256 - Because this is all inference, I would recommend connecting this inferred mechanism to the pattern that your data bear out...

L269 - the use of 'autecological' and 'synecological' strategies/interactions is used only once. I would suggest using less jargony terms or at least define them. Seems unnecessary given the terms don't appear anywhere else.

L282 - is Turnover capitalized for a reason?

L336 - would checkerboard patterns be random if there is mutual exclusivity? Seems not possible. I would expect more 'randomness' when coherence = 0? I.e. no significant structuring, which implies non-structuring, which I imagine here as somewhat random?

L342 - This implies that Turnover is only defined for coherence > 0? Is that right?

L349 - I think how 'significance' is calculated in each of these contexts needs to be described here.

L351-361 - some of these definitions and verbiage would be much more helpful in the fore-part of the manuscript.

L363 - This is really a personal decision, but because you present 3 primary metrics that form the heart of your manuscript, I would at least consider reporting how they are calculated in the materials and methods unless it is unduly complicated. This would help 'self-contain' the manuscript a bit more and make everything much easier to understand without having to bounce to other publications to gain a deeper understanding of what is going on.

References - A lot of output that is not in journal format - need curated.

Reviewer #2:

Metapopulation and metacommunity concepts have long been recognized by paleontologists (see, for example, papers in M. McKinney and J. Drake (eds.), 1998. Biodiversity Dynamics). In many ways they are an excellent model for paleontological "communities," which generally can be both spatially and temporally averaged and thus probably represent samples from a regional species pool, rather than a snapshot of a single living community. In this paper, the authors apply the metacommunity concept to Ediacaran assemblages across a wide spatial, temporal, and geographic scale. In particular, they apply the "Elements of Metacommunity Structure" framework, using an R package to analyze the data. There results suggest that the various Ediacaran assemblage show a range of metacommunity structures and are suggestive of an increase in niche specialization over time.

Overall I enjoyed reading this; I have not read about metacommunities in some time and was very pleased to be reintroduced to the concept and to see that it being actively used within paleontology. As far as I know, the use of the Elements of Metacommunity Structure and the mathematical applications of it are novel in paleontology; certainly to the study of the enigmatic Ediacaran assemblages. The figures are very nicely done. There are, however, some questions and suggestions I have before I could recommend publication:

1. Leibold et al. 2004 define "a metacommunity as a set of local communities that are linked by dispersal of multiple potentially interacting species." Given that that the Ediacaran sites within each defined assemblage ("community') are widely separated in space and time, the authors need to discuss how the metacommunity concept applies in this context. Why is it valid to treat these as metacommunities, since there is no evidence they are linked by dispersal?

2. Figure 1, and much of the results section, make no sense unless one first jumps to the Materials and Methods.

Although I understand that the journal requires that this section come at the end, that is not necessary for the discussion of the theoretical framework. The material on lines 328-361 needs to be merged with the Introduction section. The description of the data and of the R package can be kept in Materials and Methods. While this is being done, a clearer explanation of the concept of "clumped species loss" is needed; I read line 135-137 several times and I was still left confused (needed to return to the original cited papers). Is that different from "nested species loss" (caption to figure 3)?

3. In Figure 1, the central region with "start" needs to be explained in the caption. This also seems to be redrawn from the Presley et al. (2010) paper, which needs to be cited here,

4. A small point: lines 110-111, 115 "significant p-values highlighted in bold." Significant at what level? To be frank I think simply giving the p-value is sufficient; there are a great many publications that suggest we move away from the concept of "significant."

5. Lines 129-130: although there are more positive than negative associations, they are both far exceeded by the number of non-associations. This needs to be discussed.

6. Line 173-174. There is no discussion of taphonomic differences among the Ediacaran sites. This is a critical issue that needs to be considered.

7. Line 240. An incomplete sentence.

Reviewer #3:

I apologize for my delay in reviewing the manuscript due to personal issues. I read with interest the manuscript, which explores a key problem in the interface of macroevolution and ecology. Below, I list my comments on points that I think the authors need to address to improve their manuscript.

My comments:

1. Although the introduction is well written I missed a clear statement of the hypothesis and predictions the authors are testing in this manuscript. For example, Figure 1 describes what I think are hypotheses about metacommunity organization. Having said that, these hypotheses are not proper described in the introduction - the authors only describe what they want to measure but not how these measurements will be used to test predictions of distinct hypothesis.

2. On Figure 1: The laconic legend to figure makes very hard to follow it. This figure could be the heart of the manuscript but its complexity, associated with the minimalist legend imperil its usefulness for the current draft.

3. The estimating of interactions by co-occurrences is a critical assumption of the manuscript. Recently, the work by Dominique Gravel and others criticized the use of co-occurrences to estimate patterns of interaction. Although I understand the limitations of estimating patterns of interaction in fossil record, I think the authors need to at least discuss these limitations in their manuscript.

4. The use of concepts. There are a few places in the manuscript that the use of the language could be more precise and accurate. For example, in the abstract middle metacommunity may create confusion between middle referring to space or time. Accordingly, in figure 1, stable communities to the description of gleasonian metacommunities may imply these communities are less stable, what is not necessarily true. These communities are just the outcome of species-specific responses to environmental gradients. Please consider to double check the use of these terms in the manuscript.

5. Almost all positive pairwise associations are within assemblages and all negative pairwise associations are between assemblages. Assemblages, in turn, are associated with environmental factors. Thus, I would conclude that there is no need to invoke ecological interactions in shaping patterns of co-occurrence. Or, alternatively, that it is impossible to detangle the effects of environmental gradient from the effects of competition limiting occurrence of species to competitive refugia - since it is of course impossible to perform manipulative experiments. So, I think I am missing something when authors mention the role of competition structuring Ediacarian metacommunities

6. Please consider defining assemblage, community, and metacommunity in our manuscript to avoid confusion.

7. There are some descriptions of tables and figures (e.g., lines 108-111) that should be moved to legend of the table/figure.

8. Lines 142-153: Similar patterns would be predicted by multiple ecological processes. For example, Connell's intermediate perturbation hypothesis led to similar predictions. Similarly, top-down control by predators would lead to lower diversity in environments with both low and very high abundance of predators. So, my question is: is there additional evidence that this is truly a consequence of ecological succession?

9. Lines 165-167. I think I did not follow the reasoning here. If I got it then the approach used here implies that is impossible to record a positive association within this assemblage because the null hypothesis assumes 32/32 positive associations. Is that correct?

10. Line 174-179: I did not follow this flow of ideas about ecological interactions.

11. Line 240. There is something missing in this sentence, I think.

12. The discussion could be improved. There are entire paragraphs with no or just some introductory references, and I missed a truly synthesis with the literature of metacommunities, both from an evolutionary perspective (e.g., Thompson's Geographic Mosaic of Coevolution and Urban and Skelly 2006's evolving metacommunties) and from an ecological perspective (e.g., Guimaraes 2020's Annual review on ecological networks at different scales and Leibold's work on ecological processes in metacommunities).

19 Oct 2021

Submitted filename: Edenetal2021_response_final.docx

Click here for additional data file.

12 Jan 2022

Dear Dr Mitchell,

Thank you for submitting a revised version of your manuscript "Metacommunity analyses show increase in ecological specialisation throughout the Ediacaran" for consideration as a Research Article at PLOS Biology. This revised version of your manuscript has been evaluated by the PLOS Biology editors, the Academic Editor and the original reviewers. Please accept my apologies for the delay incurred while we experienced difficulties communicating with the Academic Editor over the holiday season.

IMPORTANT: You will see that while the reviewers find the manuscript somewhat improved, they still raise a significant number of concerns that must be addressed. I note a tone of exasperation in their reviews, so please do take this opportunity to make a conscientious effort to thoroughly revise the manuscript and to address these issues; we will only consult the reviewers one more time, and if they still remain dissatisfied then we may decide not to consider the manuscript further. The Academic Editor asked me to emphasise the need for sensitivity analysis: "Regarding the need for some type of sensitivity analysis, I fully agree with rev #1. The sensitivity analysis is not complex and certainly will add much more robustness to the analysis and presentation. The problem is that, somehow, the potentially confounding effects of size difference among the assemblages need to be accounted for. Rev #1 gives some cues about how to frame this."

In light of the reviews (below), we are pleased to offer you the opportunity to address the remaining points from the reviewers in a revised version that we anticipate should not take you very long. We will then assess your revised manuscript and your response to the reviewers' comments and we may consult the reviewers again.

We expect to receive your revised manuscript within 1 month.

Please email us (plosbiology@plos.org) if you have any questions or concerns, or would like to request an extension. At this stage, your manuscript remains formally under active consideration at our journal; please notify us by email if you do not intend to submit a revision so that we may end consideration of the manuscript at PLOS Biology.

**IMPORTANT - SUBMITTING YOUR REVISION**

Your revisions should address the specific points made by each reviewer. Please submit the following files along with your revised manuscript:

1. A 'Response to Reviewers' file - this should detail your responses to the editorial requests, present a point-by-point response to all of the reviewers' comments, and indicate the changes made to the manuscript.

*NOTE: In your point by point response to the reviewers, please provide the full context of each review. Do not selectively quote paragraphs or sentences to reply to. The entire set of reviewer comments should be present in full and each specific point should be responded to individually.

You should also cite any additional relevant literature that has been published since the original submission and mention any additional citations in your response.

2. In addition to a clean copy of the manuscript, please also upload a 'track-changes' version of your manuscript that specifies the edits made. This should be uploaded as a "Related" file type.

*Resubmission Checklist*

When you are ready to resubmit your revised manuscript, please refer to this resubmission checklist: https://plos.io/Biology_Checklist

To submit a revised version of your manuscript, please go to https://www.editorialmanager.com/pbiology/ and log in as an Author. Click the link labelled 'Submissions Needing Revision' where you will find your submission record.

Please make sure to read the following important policies and guidelines while preparing your revision:

*Published Peer Review*

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out. Please see here for more details:

https://blogs.plos.org/plos/2019/05/plos-journals-now-open-for-published-peer-review/

*PLOS Data Policy*

Please note that as a condition of publication PLOS' data policy (http://journals.plos.org/plosbiology/s/data-availability) requires that you make available all data used to draw the conclusions arrived at in your manuscript. If you have not already done so, you must include any data used in your manuscript either in appropriate repositories, within the body of the manuscript, or as supporting information (N.B. this includes any numerical values that were used to generate graphs, histograms etc.). For an example see here: http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1001908#s5

*Blot and Gel Data Policy*

We require the original, uncropped and minimally adjusted images supporting all blot and gel results reported in an article's figures or Supporting Information files. We will require these files before a manuscript can be accepted so please prepare them now, if you have not already uploaded them. Please carefully read our guidelines for how to prepare and upload this data: https://journals.plos.org/plosbiology/s/figures#loc-blot-and-gel-reporting-requirements

*Protocols deposition*

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Thank you again for your submission to our journal. We hope that our editorial process has been constructive thus far, and we welcome your feedback at any time. Please don't hesitate to contact us if you have any questions or comments.

Sincerely,

Roli

Roland Roberts

Senior Editor

PLOS Biology

rroberts@plos.org

*****************************************************

REVIEWERS' COMMENTS:

Reviewer #1:

General

This revision adds a lot of clarity to the initial submission, however I still have a lot of questions that I do not believe are addressed in the current version. There is still some language confusing taxa interactions vs. taxa co-occurrences that need to be addressed, and I have also tried to note areas where the text can be improved to clarify the motivation of the contribution. Specifically, I found the section detailing the author's hypothesis testing framework to be inadequate and have tried to offer suggestions on that front.

Throughout, and especially in the results, the writing needs work. There are still sentences that aren't following grammatical rules, which I feel - at this level - should have been addressed. I think that the results section could also be better organized to translate the central findings of the paper… perhaps via subheadings.

Unless I missed it completely, one outstanding question that I have with respect to the analysis as a whole is the effect of community size on the metacommunity analysis. Some of the communities vary a lot in size… Nama is very small; White Sea is very large. Are the analyses sensitive to size? If so, which are, and which are not? It's not really addressed (again, unless I missed it), except to say that some communities were not included because they were *too* small (4)… but what is too small?

Because the difference between the largest (White Sea) and smallest (Nama) is what forms the basis for interpreting whether there was either an extinction or restructuring, understanding what of the results might be influenced by the size difference is *incredibly important* in my view. Incorporating a sensitivity analysis to show that the results are insensitive to community size (hopefully) would be straightforward to do, and I think necessary to disentangle these potential effects. Or perhaps there is a) prior exploration of these effects in previous pubs using this analysis, or b) that White Sea is broken into middle/outer shelves makes the size comparisons more comparable. Yet if the latter is the argument, I would want to understand — looking at Fig 3a — what drives the differences between White Sea (all together) vs. White Sea Middle Shelf, White Sea Outer Shelf, White Sea Deep Subtidal (which look more similar to Nama)… the structure or the size?

Specific

L13 - It isn't clear in the abstract that assessing these changes in diversity using an 'Elements of Community Structure' framework is posing an alternative hypothesis to a catastrophic mass extinction

L16 - Again, the data collected cannot say anything about how taxa interact… only how they co-occur

L79-81 - is there an assumption here related to even dispersal across all localities? If there was uneven dispersal (some localities connected more strongly than others), would this change any of the assumptions in the analyses?

L84 - it should be mentioned here that these metrics are hierarchical rather than independent

L86 - From Figure 1, I only count 9… I get that the others are the Quasi/Mixed structures, but it is hard to glean from the figure. More generally I don't really understand the quasi structures…

Figure 1 - I find the orientation hard to follow… aside from the (-)(+)(0) layout (I'll let that one go), on the left negative boundary clumping is on top and positive boundary clumping is below; on the right, positive boundary clumping is on top, and negative boundary clumping is below. For the reader to keep all of these attributes in their heads, these stylistic choices can make the paper much harder to follow. There is also an errant 'minus sign' just below 'TURNOVER'… not sure if it's supposed to be there or not?

Intro - this is a nice layout - I like how the first aim is very community structure oriented, the second aim is community v environment, and the third aim is species pairwise correlations… it might be worthwhile to emphasize this tiered approach to the analysis, as I see it as a core strength of this contribution (just a suggestion)

L149 - a strange sentence: "because co-occurence does not necessarily correspond to interaction, here we interpret pair-wise interaction…" do you mean pair-wise correlations?

L153-159 - This section is strangely worded… it is stated that 3 hypotheses will be tested, but proposing a hypothesis test is different than guessing what outcome will occur from an analysis. For example, the current statement goes like this: you 1) hypothesize that increased taxonomic diversity is reflected in terms of increased taxa co-occurrences, 2) hypothesize that Ediacaran data exhibits strong metacommunity structure, and 3) will test whether there is evidence of an extinction… But in reality, you are attempting to *build support* for an alternative hypothesis for extinction… To be more rigorous about this, it seems to me that you are testing between 3 hypotheses:

1) Null: nothing happened (or the signal can't be distinguished from random)

2) Alt Hypothesis 1) an externally-caused mass extinction happened

3) Alt Hypothesis 2) an internally-caused restructuring happened.

You are proposing that a) increased ecological complexity via co-occurrences and b) strong metacommunity structure will lend support to Alt Hypothesis 3, and against 1 and 2 (I presume the Null has already been refuted in prior work showing the diversity drop, etc). Alternatively, if your metacommunity structure analysis supports negative Turnover, that would be support for Alt Hypothesis 2. If I am accurately describing the intent of this paper, it is not very well captured by the organization in Lines 153-160, so I would suggest being a bit more clear about what alternative hypotheses you are testing against, rather than stating guesses at what the results will be.

L177 - I would explicitly state what you are pooling… here it is presence/absence data of organisms as a function of locality.

L220 - I don't think the z-score is a measure of statistical significance… it is standardized deviance away from the mean… so it is telling us how similar or different each community is from the average across communities in units of standard deviation (i.e. 1 standard deviation away from the mean, etc), if I am understanding it correctly. Echoing my comment in the first draft, this needs to be clearly explained.

L228 - should this be referring to Fig 3? I don't get any insight to Coherence/Boundary Clumping from looking at Fig 2

Generally - what is the effect of size differences between communities? The Nama community is very small and the White Sea community is very big. Are measures of Metacommunity Structure sensitive to size heterogeneity? If so, how could this skew your perspective of these systems? One way to examine this… if you took random subsamples of taxa from the White Sea, where the number of subsamples is equal to that of the Nama community, measured metacommunity structure, and repeated this many many times, do you get measurement distributions with a mean similar to that using the full dataset?

L306 - strangely worded… it begins as a question but without a '?' ends as a statement.

The Results section could really use some editing… there are still quite a few sentences that are poorly constructed, and it overall feels meandering. I had a hard time understanding what the important results were relative to the less-important findings. Easy to get lost in this section. Perhaps some reorganization would help with subheadings? There was also a lot of discussion material in the results section, which I don't personally mind, though it made a discussion section feel superfluous.

L397 - Again (and again), these results cannot say anything about 'interactions between taxa'… only associations

L423-424 - This is a big leap. While there is definitely work showing that generalists may have a selective advantage in the face of large extinction events, it is quite a thing to say that the presence of generalists would imply a catastrophic extinction, or that their absence is incongruent with an extinction!

Reviewer #2:

[IMPORTANT: See attached Word file for fully formatted version, including a Table]

As a reviewer of the original submission, I am glad to have had the opportunity to provide a review of the revised manuscript of this paper. I thought the original paper had promise, but with numerous structural issues, which the other reviewers also recognized. I was pleased to see that they have attempted to address all the issues raised by the reviewers. The current manuscript is a marked improvement as a result.

That said, the current version still has numerous issues that will require at least one more round of revisions – some of these should have been caught by a careful re-reading prior to resubmission- so that is disappointing. I will try to enumerate these by line. I will also make a suggestion that I hope will improve the clarity of the paper.

1. Line 44. Is “the reduction in taxonomic diversity” different from “diversity drop”? Not clear!

2. Lines 73, 89. Both start with “First.” Reading further, the one on line 89 should be deleted.

3. Lines 73-87. Although there is evidence for larval transport and low provinciality, that does not indicate to me that these can all be treated as a metacommunity in the sense that ecologists would use it. Instead, acknowledge this but point out that for the purpose of this paper we can treat them as metacommunities.

4. Line 86-127.

a. I went back to the original paper by Presley et al. (ref. 30) and found that there were actually 14 metacommunity types, if one includes Checkerboard and Random. Each of the six categories with positive coherence and significant positive or negative turnover has a non-significant, quasi- equivalent.

b. Even with the revised figure caption and text, I was still confused about the relationships among the metrics and the community types. I was especially confused by what was meant by a “quasi-“ structure and what it implied about the metacommunity. Given that many of the analyses were consistent with a “Clementsian quasi-structure,” this needed further discussion.

c. I have made a table (see below) because it clarified for me the properties of the various metacommunity types. I suggest using something like this in the next revision.

d. I share with one of the other reviewers the issue of capitalization of the metrics. I advise not capitalizing them after they are introduced, and leave the capitalization to the metacommunity types, which are really what the paper should be about. As a comparison, “mean” and “standard deviation” are not capitalized.

e. Lines 115-124 do not belong here! They are a fragment from the results (lines 304..) (a careful readover should have caught this – do not make this the reviewers job!)

5. Lines 139-142 and Figure 2. Not an accurate description of reciprocal averaging (correspondence analysis). The method ordinates the samples (sites) by their species composition (variable); it also ordinates the species by which site they are in. Variable ordination scores are averages of the case ordination scores and case ordination scores are averages of the variable ordination scores, ”thus “reciprocal averaging.” Both plot on the same axes. Depth is an independent variable, which can then be plotted against the RA axes to help with the interpretation. So, are the axes of Figure 2 are probably the scores for the sites and species on the first RA axis; the left-hand plot of depths is independent data, not used to do the ordination.

Also: make clear in caption to Figure 2 that this a reciprocal averaging ordination.

6. Line 172. Again, there are 14, not 12 models, all six illustrated models have quasi- equivalents

7. Lines 177-186. This paragraph badly needs rewriting.

a. On line 178, you point to Figure 2, but then say nothing at all about it and then jump right away into Figure 3 and then back to Table 1. There needs to be a detailed explanations of both Figure 2 and Table 1, including what is meant by site scores.

b. Looking at the results, it is clear that the relationship between depth and scores only holds true for the Nama, so that it will strongly influence the pooled results.

8. Lines 188-190. How does the low level of non-random associations compare to what is observed in modern communities or other paleontological examples? Is this unusually low or high? See: LYONS, S. K et al.. 2016. Holocene shifts in the assembly of plant and animal communities implicate human impacts. Nature, 529, 80-83.

9. Line 211-212. What is the simulated mean value of the metric? Not discussed in the text.

10. Lines 256-259. How are the succession stages determined? Needs more detail.

11. Lines 290, 334 – this is why there needs to be a more detailed explanation of quas-structures!

12. Line 334 – so this is the same structure as the White Sea?

13. Line 395-396. Is this a significant increase, given the large number of random associations? Again, we need context in terms of other communities.

14. Line 423-426. I am confused here. Based on table 1, the White Sea, except for the outer shelf, and the Nama have the same structure. The change is from the Avalon.

15. Line 468… this is Methods and should be so indicated.

16. Lines 469-487. This is confusing; both the RA and metacommunity analyses are being discussed in the same paragraph. Break these apart and make sure the RA is carefully described.

Metrics

Community type Properties Coherence Range Turnover Boundary Clumping

Random No structure Random

Checkerboard High number mutually exclusive pairs; taxa do not respond to gradient Negative

Nested clumped Species poor communities subsets species rich communities; community synchronous response to gradient Positive Negative Positive

Nested random Species poor communities are random subsets species rich communities Positive Negative Not significant

Nested hyperdispersed Species poor communities subsets species rich communities; species respond individualistically to gradient Positive Negative Negative

Clemenstsian Community synchronous response to gradient Positive Positive Positive

Gleasonian Species respond individualistically to gradient Positive Positive Not significant

Evenly spaced Species spread along gradient with little overlap Positive Positive Negative

Quasi-nested clumped Species poor communities subsets species rich communities; community synchronous response to gradient, fewer turnover than random but not significant Positive Negative but not significant Positive

Quasi-nested hyperdispersed Species poor communities are random subsets species rich communities; fewer replacements than random but not significant Positive Negative but not significant Not significant

Quasi - nested random Species poor communities subsets species rich communities; species respond individualistically to gradient; fewer replacement than random but not significant Positive Negative but not significant Negative

Quasi-Clementsian Community synchronous response to gradient; more replacements than random but not significant Positive Positive but not significant Positive

Quasi- Gleasonian Species respond individualistically to gradient more replacements than random but not significant Positive Positive but not significant Not significant

Quasi-evenly spaced Species spread along gradient with little overlap; more replacements than random but not significant Positive Positive but not significant Negative

Submitted filename: review.docx

Click here for additional data file.

7 Mar 2022

Submitted filename: Edenetal2022_response_final.pdf

Click here for additional data file.

29 Mar 2022

Dear Dr Mitchell,

On behalf of my colleagues and the Academic Editor, Pedro Jordano, I'm pleased to say that we can in principle accept your Research Article "Metacommunity analyses show an increase in ecological specialisation throughout the Ediacaran Period" for publication in PLOS Biology, provided you address any remaining formatting and reporting issues. These will be detailed in an email that will follow this letter and that you will usually receive within 2-3 business days, during which time no action is required from you. Please note that we will not be able to formally accept your manuscript and schedule it for publication until you have completed any requested changes.

IMPORTANT:

a) You'll see that I've changed your title slightly (inserted "an" and appended "Period") for our wider readership.

b) Thanks you for providing the data and code as supplementary files and in Figshare. We will need one/both of these to be cited as the location of the underlying data in all relevant Figure legends, and I've left a note with one of my colleagues to request this.

Please take a minute to log into Editorial Manager at http://www.editorialmanager.com/pbiology/, click the "Update My Information" link at the top of the page, and update your user information to ensure an efficient production process.

PRESS: We frequently collaborate with press offices. If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximise its impact. If the press office is planning to promote your findings, we would be grateful if they could coordinate with biologypress@plos.org. If you have previously opted in to the early version process, we ask that you notify us immediately of any press plans so that we may opt out on your behalf.

We also ask that you take this opportunity to read our Embargo Policy regarding the discussion, promotion and media coverage of work that is yet to be published by PLOS. As your manuscript is not yet published, it is bound by the conditions of our Embargo Policy. Please be aware that this policy is in place both to ensure that any press coverage of your article is fully substantiated and to provide a direct link between such coverage and the published work. For full details of our Embargo Policy, please visit http://www.plos.org/about/media-inquiries/embargo-policy/.

Thank you again for choosing PLOS Biology for publication and supporting Open Access publishing. We look forward to publishing your study.

Sincerely,

Roli Roberts

Roland G Roberts, PhD

Senior Editor

PLOS Biology

rroberts@plos.org